ATI sonic anemometers

ATI manual


The following discussion all refers to Applied Technologies, Inc. sonic anemometers with the electonics in the booms. NCAR/ATD/SSSF used older-style ATI and UW anemometers with the electronics in a separate box prior to SHEBA (1997). The description below of the arrays is still applicable for these earlier arrays, but the maintenance and other mechanical issues changed. See here for some information on these earlier probes.


This probe has 3 orthogonal, non-intersecting, paths and is called a "K" probe after the initial design by Chandran Kaimal. The pathlength is 15 cm. The idea of this probe is that each path would have good exposure to the air flow which is unaffected by the other paths. Thus, the flow distortion for this array would be due simply to that of each single path, for which wind tunnel data were available. Kaimal et al. (1990) demonstrated this with a "flipped and rotated" test in the field and found residual errors in wind speed of about 2%, using a maximum 18% correction for each path.

We've taken similar data at our Marshall field site. Our best fit uses a maximum 20% correction for each path, but still reduces the overall speed error to only about 5%. This probably is due to the variation of sampling statistics as a function of height. The Kaimal data were taken on the BAO at a height of 22m and our data was taken at a height of ~5m(?). Since the Marshall data are more representative of our normal operations, we operate the K-probes with a maximum 20% single path correction but with no further correction for flow distortion effects from the array.

NUW probe

This probe is built by NCAR, based on a design initially conceived of by Joost Businger and constructed at the University of Washington (Zhang et al., 1986) -- hence the name "New U.W.", or "NUW". The idea of this probe is to have no objects upwind of a path, including the transducer itself, which requires that no path be oriented horizontally. Of course, one "strut" must be used to physically connect the top and bottom parts of the array, so one wind direction must considered "bad". Fortunately, flow from this direction also goes through the tower that the sensor is mounted on, so only this direction must be ignored. A pathlength of 20 cm was used in the original UW probes and was duplicated in the NUW probes to allow the earlier flow distortion tests to be directly applied.

This probe has been tested in a large EPA wind tunnel and in the field and found to have minimal flow distortion, except when the wind direction is within 20 degrees of the back of the array. Thus, we operate it with no corrections for flow distortion and reject data from directions corresponding to flow from the tower.

The NUW arrays are not perfectly built, despite the use of a jig when they are epoxied together. Thus, the array geometry has to be measured whenever the arrays are rebuilt. I have found that using spherical balls with a precisely-drilled hole for the transducer to fit inside provides a good reference surface to allow the dimensions to be obtained using a larger vernier caliper. From a measurement of all combinations of distances between transducers, it is possible to get the angles to within ~0.2 degrees and an estimate of the closure of the measurements (generally to ~2mm).


This is a new "ring" sonic developed by ATI. The electronics are identical to a K-probe. Like the K-probe, the pathlength is 15 cm. Like the Gill, CSAT3, and UW probes, it has 3 paths that are separated by 120 degrees in azimuth and pitch angles all 60 degrees. Since one path is oriented perpendicular to the boom, this array is symmetric to being flipped upside-down (which also makes it cheaper to manufacture, since the same piece can be used for the top and bottom of the array). ATI extended their operating code to apply the appropriate matrix rotations (and path curvature corrections, I think) to deal with this non-orthogonal array, so our NUW sonics also use this code.

We have inadvertently taken data with the NUW probes using the rotation matrix for an A probe, so its rotation matrix is recorded here.

Level sensors

ATI offers an option to add level sensors to their sonic anemometers to help in orienting the sensor and/or data. We had a long discussion on whether to add these to all of our probes (initially we only had 2 arrays with levels) and decided not to add level sensors to the rest of our arrays. For the data from those two sensors, a description of how to interpret the data is given here.

Spike detection

The current generation of ATI sonics have significantly improved the quality of the signal, so that spikes in the data are much less of a problem. The ATI electronics have spike detection built in (see description), but we have found that temperature jumps due to daytime convection in the atmospheric surface layer gave false detections. Thus, we currently turn this internal spike detection off.

However, we do process the data using a more sophisticated despiking algorithm, which looks at the 10 or 20 sps data from the sonic. In most cases, we use the output of this algorithm just to flag potentially bad data, though we have implemented the algorithm to allow us to replace these values as well.


I have a checklist of how to service these anemometers, including transducer replacement, NUW array reconstruction, calibration, and software reconfiguration. Any major maintenance should be logged in Service.txt, so that we can keep track of the history of each sensor.

To check the ATI sonic on the bench, I have a view of the oscilliscope display for a normal ATI sonic when the oscilliscope probes are connected to test points on the ATI electronics board, horizontal sweep at 0.5ms/div and vertical scale at 2V/div. This shows the U1 transducer sending an accoustic pulse which is received about 700us later by U2 and then a wait of about 200us before V1 transmits to V2, etc. This cycle repeats every 5 ms. Note that the amplitude of the received signal is about 5 V_half-peak. If it is significantly lower, the transmit and/or receive transducer should be replaced. Also note that a 'scope with delayed sweep capability will allow you to look at the pulse shapes in detail.


Kaimal, J.C., J.E. Gaynor, H.A. Zimmerman, and G.A. Zimmerman, 1990, "Minimizing flow distortion errors in a sonic anemometer", Boundary-Layer Meteor., 53, 103-115.

Zhang, S.F, J.C. Wyngaard, J.A. Businger, and S.P. Oncley, 1986, "Response characteristics of the U. W. sonic anemometer", J. Atmos. Ocean. Tech., 3, 315-323.